According to statistics, the accident rate caused by fatigue driving and mental inconcentration in traffic accidents accounts for 93%. To solve this problem, 68 key points of face are introduced in this paper. Based on deep learning and machine vision algorithm concepts, combined with YOLOv8 model and related fatigue driving judgment mechanism, PERCLOS algorithm, MAR algorithm, EAR algorithm and HPE algorithm are used to improve the accuracy and reliability of the system. A set of algorithms for recognizing tired driving behavior is successfully constructed.

Helmet detection is a computer vision task with important application value, involving safety management in many fields such as construction sites, mines, and electric power. However, helmet detection also faces many challenges, such as large changes in target size and aspect ratio, rapid changes in target velocity, target occlusion, and background interference. In order to solve these problems, this paper proposes a safety helmet detection method based on YOLOv8, which uses the characteristics of high speed and high precision of YOLOv8 combined with the characteristics of safety helmets to achieve effective detection and identification of safety helmets.

In order to alleviate the problem of seat tension in university canteens, it is urgent to develop a kind of software to solve this problem and improve the dining experience for students. This article discusses the use of the current mainstream microservice framework Spring Boot, combined with technologies such as MyBatis Plus and Vue.js, to design and implement a campus ordering system backend, which facilitates platform administrators and merchant administrators to manage dishes, packages, orders, etc., providing a solution for the problem of tight dining for college students.

With the rapid development of information technology, computer network security has become a focus of social attention. This article explores the teaching innovation and practice of Computer Network Security from the perspective of "ideological and po‐ litical guidance, integration of industry and education", aiming to cultivate outstanding talents with high awareness and skills of net‐ work security. The course utilizes Xuexi Tong APP to carry out blended online and offline teaching. Through the reconstruction of course content, improvement of teaching modes, innovation of teaching methods, integration of ideological and political elements, reform of teaching evaluation, and promotion of teaching effectiveness, the educational philosophy of "moral education first, studentcentered, integration of industry and education, and continuous improvement" is adopted to achieve the deep exploration of ideologi‐ cal and political elements in the course while imparting professional knowledge, integrate professional knowledge of network and se‐ curity with production practice, strengthen school enterprise cooperation, continue teaching reform, and build a learning community for teachers and students.

With the rapid development of information technology and the growing demand for financial management and financial analyses, bookkeeping is no longer limited to ordinary handwriting or manual recording in software. In order to meet the needs and development of today's era, an intelligent financial system based on OCR and large language model is designed. The system is suit‐ able for mobile phones, with financial management functions as the starting point and simplified management operations as the pur‐ pose, and realizes a system that is mainly based on OCR recognition scanning and recording bookkeeping functions, supplemented by financial analysis functions of big language model, big data consumption classification trends, and multi-version switching of ver‐ sions suitable for people of different age groups. The system not only greatly improves the efficiency of people's bookkeeping, but also promotes the intelligent financial transformation of society.